To enhance its accessibility during later stages of development, the chicken embryo can be cultured in a dish without adverse effects on development 32333435. To have access to the developing cerebellum, we transferred embryos to a domed dish after the second day of incubation (Additional file 1). As described previously 35, the transfer should be done no later than embryonic day 2.5 (E2.5) for best survival of the embryo. The survival rate of embryos depended on several factors, including temperature and humidity in the incubator, as well as the time of embryo transfer. Routinely, we found 62% of the embryos alive on E8, the time of injection in this study. At that time, embryos grown in a dish did not differ from embryos kept in the egg. Similarly, the comparison of cerebellar development between embryos developing in the egg and those developing in a dish for up to ten days after transfer (the latest time point that we analyzed) revealed no differences (Additional file 2).

Injections into the cerebellum had to be carried out without direct visual control. However, the blood vessels that are readily visible through the skin could be used as landmarks (Figure 1a). By using an expression plasmid encoding enhanced green fluorescent protein (EGFP), the proper injection site, the required depth of the injection, as well as the best electroporation settings were determined. For electroporations we placed the head of the embryo between tweezer electrodes. To prevent tissue damage, direct contact between embryo and electrodes was carefully avoided (Figure 1b). The cerebellum was efficiently transfected with 6 pulses of 99 ms duration at 40 V, as judged by the number of EGFP-expressing cells (Figure 1d–i). Only one half of the cerebellum expressed the transgene, because the negatively charged nucleic acids were migrating towards the anode (Figure 1c,d). Therefore, the untransfected side of the cerebellum could be used as a control for the analysis of potential phenotypes. When injections were carried out at Hamburger and Hamilton stage 34 (HH34; E8) 36, the survival rate one day after the experiment was 72%. However, the time point of the injection is not restricted to HH34. Injections and electroporations were performed successfully until HH36 (E10; not shown).

Depending on the depth of the injection, different cerebellar layers could be targeted. Injections into the ventricle resulted in transfection of the ventricular zone, the origin of Purkinje cells and interneurons (Figure 1e). After superficial injections into the cerebellar anlage, cells of the developing ML (Figure 1f,g) and granule cells in the EGL were transfected (Figure 1h,i), similar to observations made in postnatal rat pups 37. To get efficient transfection of all cerebellar layers, the glass capillary was inserted into the ventricle and the injection pressure was maintained during its retraction. For this study, we used only sections that were successfully transfected in the EGL and the developing ML. For analyses of experimental and control embryos, we carefully matched cerebellar sections with respect to their anteroposterior and mediolateral positions.

To make sure that injection and subsequent electroporation of plasmids encoding EGFP or of long double-stranded RNA (dsRNA) into the cerebellum did not induce non-specific effects, we checked for apoptosis and histological changes in the transfected area of the cerebellum (Figure 2). Apoptosis was compared between untreated control embryos (Figure 2a), control embryos injected and electroporated with the EGFP plasmid alone (Figure 2b), and experimental embryos injected and electroporated with EGFP and dsRNA derived from AX-1 (Figure 2c). We did not find any difference in apoptosis between control and experimental embryos, indicating that our experimental procedure did not induce cell death. This is consistent with our previous findings demonstrating that injection and electroporation of long dsRNA did not induce unspecific effects in the chicken spinal cord 25.

The comparison of sections from control-injected/electroporated and untreated control embryos stained with methylene blue confirmed the absence of morphological changes in experimental embryos (not shown). The cerebellar foliation and the organization of the EGL were unchanged three days after electroporation. Furthermore, the organization of Purkinje cells and Bergmann glia cells was visualized by staining for Calbindin (HH38; Figure 2d–f) and Vimentin (HH37; Figure 2g–i), respectively. No changes were observed when experimental embryos were compared with control embryos three and four days after electroporation, respectively. Taken together, we did not find any evidence that ex ovo RNAi induced apoptosis, morphological changes, or developmental delays during cerebellar development.

AX-1 is expressed in postmitotic granule cells in the EGL of the developing cerebellum (678; this study). To demonstrate efficient downregulation of AX-1 in these cells by ex ovo RNAi, we mixed dsRNA derived from AX-1 with a plasmid encoding EGFP at a molecular ratio of 50:1. In all experiments EGFP expression was used to identify the position and the size of the electroporated area in the cerebellum. Electroporation of AX-1 dsRNA efficiently knocked down AX-1 protein (Figure 3). Caudal cerebellar sections taken from an experimental embryo exhibited a patchy pattern of AX-1 expression (Figure 3b) in contrast to the homogenous appearance in a control embryo (Figure 3d). As expected, the vast majority of EGFP-expressing cells did not express AX-1 (Figure 3c). Consistent with the 50-fold excess of dsRNA molecules compared to the EGFP plasmid, we found many cells that no longer expressed AX-1 but failed to express EGFP.

For the quantification of gene silencing we calculated the relative intensity of AX-1 staining in the electroporated area (Figure 3f) compared to the corresponding non-targeted area of the cerebellum (Figure 3g). A strong decrease of more than 60% of the AX-1 staining intensity was seen in the electroporated area of the cerebellum (identified by EGFP expression; Figure 3e) when embryos were injected and electroporated with AX-1 dsRNA (Figure 3i). As expected, there was no downregulation of AX-1 in the transfected area of embryos injected with the EGFP plasmid alone (Figure 3h,i).

In order to demonstrate the specificity of ex ovo RNAi, we monitored the expression of the related cell adhesion molecules NgCAM, NrCAM, and Contactin/F11 that are expressed in the developing chicken cerebellum (Figure 4). In agreement with published reports, we found that AX-1 preceded NgCAM, NrCAM, and Contactin/F11 expression in postmitotic granule cells 383940. After downregulation of AX-1 no changes in the expression of NgCAM, NrCAM, and Contactin/F11 were observed (Figure 5). Taken together, these results indicate that ex ovo RNAi with long dsRNA derived from AX-1 effectively knocked down AX-1 levels without affecting non-targeted but related genes expressed in the cerebellum.

For functional analysis of AX-1 in the developing cerebellum, we injected two long dsRNA fragments that were used previously to analyze the function of AX-1 in commissural axon guidance by in ovo RNAi 25. Injection of either one of the dsRNAs derived from AX-1 resulted in the same respective phenotypes in the spinal cord and in the cerebellum. In the spinal cord, loss of AX-1 caused the failure of commissural axons to cross the midline 3425. Knock down of AX-1 in the developing cerebellum at HH34 resulted in the failure of granule cell axons to extend parallel to each other and parallel to the pial surface (Figure 6). When analyzed in coronal slices at HH35, parallel fibers were found to deviate from their normal trajectory, often even invading the outermost part of the EGL, where granule cell precursors proliferate (Figure 6c,f). In control embryos injected with dsRNA derived from CD34, a gene not expressed in the cerebellum, or in control embryos injected with the EGFP plasmid alone (Figure 6e), neurofilament staining was not found in the outer EGL. The same was true for the contralateral side of experimental embryos (Figure 6d). The failure of granule cell axons to extend in a parallel manner with respect to each other and with respect to the pial surface was reflected in the irregular appearance of the ML (Figure 6b). Its width was non-uniform and the fiber density appeared strongly reduced. Because we did not observe an increase in cell death compared to control embryos (Figure 2) we took the apparent decrease in fiber number in the developing ML as further evidence for the aberrant growth of parallel fibers. Because it was impossible to count the number of aberrant fibers individually, we measured the intensity of neurofilament staining in the developing ML. As the width of the layer varies considerably depending on the anteroposterior position in the cerebellum, we measured the staining intensity as a ratio between the targeted and the contralateral side. The comparison of neurofilament staining intensity between the experimental and the control sides indicated a 38% decrease in AX-1 dsRNA-treated embryos (n = 19 slices from 9 embryos) but no changes in EGFP-treated embryos (n = 14 slices from 6 embryos) or untreated embryos (n = 10 slices from 4 embryos) (Figure 6i).

These results were reproduced qualitatively in embryos treated with function-blocking anti-AX-1 antibodies (Figure 6g,h). Thus, perturbation of AX-1 function either by ex ovo RNAi or by injection of function-blocking antibodies resulted in aberrant growth of granule cell axons.

Loss of AX-1 in vivo did not prevent the formation of granule cell axons, but interfered with their orientation (Figure 7). Previous in vitro studies did not detect an effect of TAG-1/AX-1 on mouse granule cell axon extension 41. Consistent with these results, we found that blocking AX-1 function in vivo did not interfere with axon extension but with the direction of their growth. The comparison of granule cells in slices taken from control-injected embryos and embryos lacking AX-1 at high magnification revealed normal T-shaped axons in controls at HH35 (Figure 7a) but aberrant growth of axons toward the pial surface in experimental embryos (Figure 7b,c).

To rule out an effect of aberrant granule cell development on parallel fiber formation, we analyzed the migration of granule cells from the EGL through the developing ML. For this purpose we used Pax6 as a marker for granule cell somas 42 at HH35 (Figure 8a,b) and HH38 (Figure 8c,d). Downregulation of AX-1 did not interfere with granule cell migration (Figure 8b,d) compared to untreated controls (Figure 8a,c). Furthermore, the lack of AX-1 did not interfere with granule cell maturation. In particular, the number and position of proliferating granule cell precursors in the outer EGL was indistinguishable between control embryos (Figure 8e) and experimental embryos (Figure 8f). Thus, lack of AX-1 did not interfere with granule cell proliferation and migration.

To address the mechanism underlying the observed failure of granule cell axons to extend in a parallel manner to each other and to the pial surface, we investigated the role of AX-1 in neurite outgrowth promotion in vitro. When laminin was used as a substrate, granule cell axons grew almost exclusively on or in close contact with glia cells. Blocking AX-1 function did not interfere with this behavior nor change neurite length (Figure 9). Neurite length was 101.0 ± 7.3 μm for granule cells grown without antibodies added to the medium, 101 ± 9.7 μm in the presence of non-immune goat IgG, and 94.5 ± 7.6 μm in the presence of goat anti-AX-1 IgG.

Further confirmation for the absence of an effect of AX-1 on granule cell axon outgrowth was found when we coated culture dishes with purified AX-1. As described earlier 4344, AX-1 supported axon outgrowth and the formation of characteristic, large growth cones from dorsal root ganglion neurons. However, under the same conditions, we did not detect neurite outgrowth from granule cells (Figure 10), confirming previous results obtained with mouse granule cells and TAG-1 41. In agreement with our previous observations of granule cells cultured in the presence of function-blocking antibodies (Figure 9), they extended neurites on laminin, no matter whether they expressed AX-1 or not, that is, when AX-1 was downregulated by in vitro electroporation before plating (Figure 10g–l). In contrast, no axons were found on AX-1 used as a substrate (Figure 10a–f). Therefore, we concluded that AX-1 was required for axon guidance but not for axon growth, in agreement with earlier observations made for commissural neurons from the dorsolateral spinal cord 45.

Fertilized Hisex eggs were obtained from a local hatchery and pre-incubated for 2.5 days at 38.5°C with a humidity of at least 45%. After the egg had been positioned on the side for at least 20 minutes to allow for the embryo to float on top of the yolk, it was wiped with 70% ethanol, and cracked on a sharp edge. The whole egg content was transferred carefully into a domed dish with a diameter of 80 mm and a depth of 40 mm. These dishes were produced for the food industry from oriented polystyrene (OPS; Bellaplast, Altstaetten, Switzerland; Additional file 1). To minimize evaporation, the dish was covered with a lid. The ex ovo cultures were kept in an egg incubator at 38.5°C (Heraeus B12, Therrmo Fisher Scientific, Wohlen, Switzerland). To evaluate the survival rate at different stages of development, we opened the incubator once a day to count survivors (Additional file 1d). However, for routine experiments and best survival rate the incubator was not opened for the first 6 days of ex ovo culturing. Under these conditions the survival rate was much better (on average 62% at E8). Compared to the ex ovo culturing method described by Luo and Redies 35, the use of a domed dish was advantageous. We had better survival for a longer period of time in the domed polystyrene dishes. Contaminations were very rare (much less than 1%), although we did not work in a laminar flow hood.

For in vitro transcription, 2 μg of the linearized (cut with BamHI and SacI; Roche, Basel, Switzerland) and purified plasmids encoding AX-1 (either nucleotides 28–1,592 or nucleotides 1,620–2,298 of the full-length cDNA; see 25 for details) were mixed with 0.8 μl 100 mM rNTPs (25 mM each; Roche), 0.5 μl RNasin (Promega, Duebendorf, Switzerland), 2 μl Sp6 or T7 RNA polymerase (15 U/μl; Promega), and 2 μl 100 mM DTT in transcription buffer (final volume 20 μl). After 4 h at 37°C, the DNA template was removed from the in vitro transcription mixture by digestion with 20 units RNase-free DNaseI (Roche) for 1 h at 37°C. For dsRNA derived from CD34, nucleotides 1,640–2,417 of the full-length cDNA (XM_417984) were used. The plasmid was linearized with NotI (New England Biolabs, Ipswich, MA, USA) and EcoRI (Roche) and single-stranded RNA (ssRNA) was prepared using T3 and T7 RNA polymerases. The synthesized ssRNA was purified by sequential extraction with acidic phenol-chloroform (25:24:1 vol/vol/vol phenol/chloroform/isoamyl alcohol) and chloroform/isoamylalcohol (24:1). After precipitation with 100% ethanol, the ssRNA was washed with 70% ethanol, and dissolved in 20 μl phosphate-buffered saline (PBS). To produce dsRNA, equal nanogram amounts of antisense and sense ssRNAs were mixed, heated for five minutes at 95°C, and then slowly cooled down to room temperature by switching off the heating block. For quality control, 1 μl samples were taken after each step and analyzed by gel electrophoresis. The dsRNA was stored at -80°C until further use.

Injections and electroporations were performed at E8. The embryos were staged according to Hamburger and Hamilton 36. To have direct access to the embryo, a small hole of 3–4 mm diameter was cut into the extraembryonic membranes above the eye. For positioning and stabilization of the head during injection and subsequent electroporation, we used a hook prepared from a spatula (Figure 1a). Approximately 1 μl of the nucleic acid mixture consisting of a plasmid encoding EGFP under the control of the β-actin promoter (100 ng/μl), dsRNA derived from AX-1 (500 ng/μl), and 0.04% (vol/vol) Trypan Blue (Invitrogen, Carlsbad, CA, USA) dissolved in sterile PBS were injected into the cerebellum using a borosilicate glass capillary with a tip diameter of 5 μm (Figure 1a; World Precision Instruments, Berlin, Germany). Depending on the depth of the injection, different cerebellar layers could be targeted. To get transfection of all cerebellar layers, the glass capillary was first inserted into the ventricular system and injection pressure was maintained during retraction. Before electroporation a few drops of sterile PBS were added to the embryo. For the electroporation a platelet electrode of 7 mm diameter (Tweezertrodes Model #520, BTX Instrument Division, Harvard Apparatus, Holliston, MA, USA) was placed collaterally to the head of the embryo (Figure 1b). Six pulses of 40 V and 99 ms duration with one second interpulse intervals were applied using a square wave electroporator (ECM830, BTX). For the present study, we selected only embryos that were successfully electroporated in the EGL and the developing ML for further analysis. Slices and sections from experimental embryos were compared with slices/sections taken from the same position of a control cerebellum with respect to both the anteroposterior and the mediolateral axis. As controls we used CD34 dsRNA injected, EGFP plasmid injected, and untreated control embryos, respectively.

For injections of function-blocking anti-AX-1 antibodies, approximately 1 μl of the antibody solution (10 mg/ml with 0.04% Trypan Blue in PBS) was injected into the cerebellum three times every 12 h.

One to four days after electroporation the embryos were sacrificed. The brain was removed and analyzed for EGFP expression using a fluorescence stereomicroscope (SZX12, Olympus). After fixation for 2 h at room temperature in 4% paraformaldehyde in PBS, the brain tissue was rinsed in PBS and transferred to 25% sucrose in 0.1 M sodium phosphate buffer, pH 7.4, for cryoprotection. In this study, 30-μm-thick sagittal cryostat sections and 250-μm-thick coronal vibratome slices were used. For the preparation of cryostat sections, the brains were embedded with OCT Tissue-Tek in Peel-a-Way^® disposable embedding molds (Polysciences, Eppelheim, Germany), frozen in isopentane on dry ice, and cut on a cryostat (Leica CM1850). The sections were collected on SuperFrost^®Plus microscope slides. For the preparation of 250-μm-thick vibratome slices, the brain was embedded in 6.5% ultra low-melting agarose (Type IX, Sigma, Buchs, Switzerland) and cut with an OTS-300-04 oscillating tissue slicer (Electron Microscopy Sciences, Hatfield, PA, USA). The slices were collected in PBS and the remaining agarose was removed before staining.

Immunostaining of cryostat sections was done essentially as described earlier 5. We used the following primary antibodies: rabbit anti-GFP (1:250; Abcam, Cambridge, UK), FITC-labeled goat anti-GFP 1:400; Rockland, Gilbertsville, PA, USA), anti-Calbindin (1:1000; Swant, Bellinzona, Switzerland), both rabbit and goat anti-AX-1 (1:1,000) 325, goat anti-NgCAM, anti-NrCAM (both 1:1,000), rabbit anti-Contactin/F11 (1:1,000) 5, mouse anti-neurofilament (RMO270; 1:1,500; Zymed/Invitrogen, Carlsbad, CA, USA), rabbit anti-neurofilament (1:250; Chemicon/Millipore, Billerica, MA, USA), and the monoclonal antibodies H5, recognizing Vimentin (supernatant), and Pax6 (2 μg/ml; both obtained from the Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA). All antibodies were diluted in blocking buffer (10% fetal calf serum in PBS). Incubation was overnight at 4°C for primary antibodies and 90 minutes at room temperature for secondary antibodies. Secondary antibodies were: goat anti-rabbit-Alexa488 and donkey anti-goat-Alexa488 (1:250; Molecular Probes/Invitrogen), goat anti-mouse-Cy3 (1:250), donkey anti-rabbit-Cy3 (1:200), or donkey anti-goat-Cy3 (1:250; all Jackson ImmunoResearch Laboratories, Newmarket, Suffolk, UK). For staining of vibratome slices, essentially the same protocol was used, although incubation times were extended 72. Slices were mounted in sterile PBS between two coverslips sealed with high vacuum grease (Dow Corning, Wiesbaden, Germany).

Embryos were injected with dsRNA derived from AX-1 and the EGFP plasmid, or with the EGFP plasmid alone, and electroporated at HH34. After 4 days (HH38), 200 μl 50 mM bromodeoxyuridine (BrdU) in H₂O were pipetted onto the chorioallantois. After 3 h the embryos were sacrificed, and the brains were dissected and prepared for cryostat sections as described above. For visualization of the incorporated BrdU, the sections were incubated in 50% formamide, 0.3 M NaCl in 0.03 M tri-sodium citrate buffer, pH 7.0, for 1–2 h at 65°C, rinsed twice with 0.3 M NaCl in 0.03 M tri-sodium citrate buffer for 15 minutes, followed by incubation in 2 N HCl for 30 minutes at 37°C. Sections were rinsed in 0.1 M borate buffer (pH 8.5) for 10 minutes at room temperature, followed by PBS (six changes). BrdU was detected using the mouse anti-BrdU antibody from Sigma (1:200) using the protocol detailed above. Sections were counterstained with DAPI (5 μg/ml in PBS) for 20 minutes at room temperature.

Apoptosis was compared between embryos taken directly from the egg, control embryos that were injected and electroporated with the EGFP plasmid alone, or experimental embryos injected with dsRNA derived from AX-1 (together with the EGFP plasmid). As a positive control, sections were treated with DNase I (300 U/ml; Roche) for 10 minutes at room temperature. To detect apoptotic cells, the ApoAlert DNA Fragmentation Assay Kit (Clontech, Mountain View, CA, USA) was used according to the manufacturer's instructions. The fragmented, fluorescein-labeled DNA was detected with an alkaline phosphatase-conjugated sheep anti-FITC antibody (1:1,000; Roche) dissolved in 10% fetal calf serum/PBS. The alkaline phosphatase reaction was carried out as described earlier 73.

Fluorescence intensities after AX-1 or neurofilament staining were measured with the analySIS Five software from Olympus Soft Imaging System (Muenster, Germany). A decrease of the protein level was calculated from the relative fluorescence intensities measured in the ML of the transfected versus non-transfected areas. The staining intensity was then compared between embryos injected with dsRNA derived from AX-1, embryos injected with the EGFP plasmid only, and untreated controls. For statistical analyses ANOVA with Bonferroni correction was used. Values are given as mean ± standard error of the mean. P-values were < 0.0001.

Granule cells were collected by microdissection of the cerebellum at HH36. The EGL was separated from coronal slices by cutting through the ML. The tissue was collected in ice-cold HBSS (Hank's Balanced Salt Solution). After trypsinization and trituration, cells were suspended in MEM containing 10% fetal calf serum and penicilin/streptomycin, and plated at a density of 50,000 cells per well of an 8-well LabTek slide. Slides were coated with 20 μg/ml laminin or 50 μg/ml AX-1 essentially as described in 44 except that slides coated with laminin were precoated with 100 μg/ml poly-lysine. After 48 h, cells were fixed in 4% paraformaldehyde for 1 h at room temperature. Cell cultures contained granule cells and Bergmann glia cells when coated on laminin as determined by Pax6 and GFAP staining (not shown). On AX-1 substratum glia cells were not able to spread. Neurite lengths were measured and are representative for four different cultures. Antibodies were added to the culture medium at a concentration of 5 μg/ml.

Granule cells suspended in culture medium (see above), 40 ng/μl EGFP plasmid, and 5 ng/μl dsRNA derived from AX-1 were electroporated using the BTX Safety Stand 630B and matching cuvettes connected to the BTX Electro Square Porator ECM830. After electroporation with 1 pulse of 5 ms duration at 112 V cells were kept at 37°C for 5 minutes before plating. When cells were subject to electroporation before culturing they were plated at a density of 800,000 per well.